Here is the Eliezer Yudkowsky paper, “MIRI announces new ‘Death With Dignity’ strategy”, mentioned in the talk.
Yet to watch the video, but a quick check on the linked article and the author, left the following impressions:
- An auto-didact “social scientist”:
Eliezer Yudkowsky - Wikipedia
Yudkowsky is an autodidact and did not attend high school or college. He was raised as a Modern Orthodox Jew.
- Producing “word salad” with scientific-sounding terms:
will at best let somebody at DeepMind give a highly speculative warning about how the current set of enormous inscrutable tensors, inside a system that was recompiled three weeks ago and has now been training by gradient descent for 20 days, might possibly be planning to start trying to deceive its operators.
He likely has no idea how bad gradient descend is when it comes to optimization of real world problems.
Thiel would probably sacrifice his entire fortune if he knew it would guarantee a revolution in “social science” – as something like Ockham’s Guillotine quite plausibly could. As I’ve pointed out on numerous occasions here, Yudowsky’s “LessWrong” has played no small part in occupying the position in the “singularity” space that should have convinced Thiel, as early as the advent of the Hutter Prize in 2006, lossless compression incentives of wide-ranging longitudinal social measures stands a significant chance of catalyzing just such a revolution. Not worthy of his entire fortune but certainly worthy of a significant chunk of the Thiel Foundation’s endowment.
It’s really crazy that there is no infrastructure for this kind of incremental prize competition that would also permit a revival of the Methuselah Mouse Prize as well as the Hutter Prize and Ockham’s Guillotine – not from Kaggle, X-Prize Foundation, Patreon, Paypal, etc… nothing. (And, yes, I’ve asked.)